Nathan Wan

Nathan Wan is an artificial intelligence leader whose career spans Google speech recognition, blood-based cancer diagnostics, microbiome drug discovery, and healthcare revenue-cycle management. As head of AI at Ensemble Health Partners in 2025, he focused on a consequential healthcare problem often overshadowed by clinical applications: preventing insurance denials and helping hospitals recover payment for necessary care.

Wan studied at Carnegie Mellon University and began his career at Google, building software for operational teams and working on speech recognition and language modeling. He contributed to early efforts to transcribe doctor-patient conversations, anticipating today’s ambient clinical-documentation systems. His research on clinical speech recognition addressed challenges including overlapping speakers, inconsistent audio, and specialized medical terminology.

At Freenome, Wan helped build models and teams for blood-based cancer detection. He was co-first author of a 2019 colorectal cancer study investigating whether machine learning applied to plasma cell-free DNA could identify early-stage disease. The research highlighted a central risk in biological modeling: algorithms can mistake differences in patient populations, collection sites, or laboratory processing for genuine disease signals.

He pursued that problem through research on confounder-resistant biological models and METCC, a metric-learning approach designed to distinguish meaningful biological differences from technical variation. He subsequently served as chief technology officer at Nuanced Health, working on complex microbiome data and drug discovery, and helped organize the Los Angeles Bits in Bio community.

• Prevent insurance denials upstream. Wan connects historical claims outcomes with registration, prior-authorization, and procedural data to identify missing information or inconsistent payer requirements before a claim fails. His healthcare revenue-cycle work treats avoidable denials as coordination failures across an entire patient journey.

• Build confounder-aware biological models. His diagnostics research emphasizes validation against sequencing batches, institutions, patient characteristics, and other variables that can inflate apparent performance without capturing underlying biology.

• Preserve clinical accountability in AI-assisted appeals. Wan combines electronic medical records, clinical guidelines, and payer policies to prepare clinician-reviewed denial appeals. Medical experts retain final responsibility because generic generated letters cannot reliably satisfy organization-specific quality standards or complex medical-necessity requirements.

• Evaluate operational results, not generated text. He assesses healthcare AI using measurable outcomes including turnaround time, denial overturn rates, and workflow capacity, linking model performance to whether providers recover appropriate payment more efficiently.